Capabilities and availability¶
Start with the task that fits your environment. The complete product journey describes the destination; this page helps you choose what to use now.
Availability reviewed on 2026-10-17. The public PyPI release checked is
0.0.67; the reviewed source checkout is 513e8e08 (0.0.68.dev777). Source
features can be ahead of the distributed package. Check your installation with
python -m pip show pycharter and the relevant command's --help.
The landing update through source 9b758a09 also qualifies the source
Pydantic floor (>=2.11.0,<3.0) and adds minimum/current installed-wheel checks
for the README's local examples. Published-release and optional-service
qualification remain separate from those source checks.
| Capability | Current starting point | Scope |
|---|---|---|
| Local record validation | First-result guide | Runnable on the public release without a server, database or credentials |
| File-based pipeline enforcement | Local data feed | Explicit step contract; file output, repeat and deliberate-failure exercise |
| Contract authoring and exchange | Contracts, ODCS | Supported schema/rule/codec subsets apply; importing a rule does not guarantee a native executor |
| Python, CLI, HTTP and UI | API references, API/UI guide | Existing interfaces; complete task and result parity is being finished. Saving a configuration does not prove an executed delivery |
| MCP and record-verification agents | Agent guide, MCP setup | Tool exposure depends on policy, transport and installed version. Hosted operation requires its configured identity and storage |
| Bounded FMP delivery | Reference application | Reviewed source checkout; bounded EOD scope, synthetic fixtures and destination receipts. Exact bundle distribution and live deployment have separate gates |
| Streaming and bridge | Streaming overview | Experimental; broker, sink and delivery-mode guarantees must match the selected profile |
| Agent-acquired pipelines | Agent-step boundary | Configuration and acquisition foundations exist in source; browser runtime and end-to-end delivery remain under qualification |
| Seamless verify → deliver → schedule → recover | Product journey | Intended integrated experience; first-run UI, cross-interface handoff and operational profiles are still being completed |
Choose an operating mode¶
Embedded validation: run the library in your application. Pure record checks need no registry or always-on service.
Scheduled batch: run PyCharter inside your existing scheduler's job. The selected profile declares its destination, durable progress/evidence and independent monitor. A PyCharter API, UI or queue worker is optional for this mode.
Streaming: operate the selected broker and long-running worker. Treat replay, acknowledgement and delivery guarantees as profile-specific.
Shared service: operate the API/UI or remote MCP endpoint with authentication, resource permissions, secret handling and lifecycle policies. Production qualification of the complete shared experience is in progress; a local demo is not evidence of service isolation or recovery guarantees.
Interpret results precisely¶
Contract conformance means the checked data satisfied the declared requirements. It does not establish completeness without coverage requirements, factual truth without suitable evidence, or publication without a destination outcome.
Likewise, a saved pipeline is a definition; a verified sample is a bounded check; a successful deployment needs runtime evidence. The interface and examples must state which result they establish.
Keep installation and documentation aligned¶
Use Installation to select extras. The bounded source references identify their own prerequisites; a plugin or MCP connection does not install those applications or provision a scheduler. For a released project, pin its package version and keep the matching contract/pipeline assets with it.